Filtered Tensor Ring-Based Algorithm for Low-Rank Image Completion

The issue of image completion has attracted increasing interest in many areas of science and engineering, including smart and information technologies. In this study, we propose a new computational method for a low-rank image completion problem, where an incomplete image has a large number of uniformly distributed missing pixels or a regular missing structure, e.g. a grid. The proposed method combines a low-rank approximation, enforced by the tensor ring decomposition model, with in-line low-pass filtration. The numerical experiments with strongly incomplete images demonstrate high efficiently of the proposed method.